More formal definition of array index

[annevk]

https://tc39.github.io/ecma262/#sec-array-exotic-objects seems to define array index somewhat informatively. Could there be a more formal definition there so it can be reused by IDL?

You probably also do not want to define it only for String properties, since as far as I can tell https://tc39.github.io/ecma262/#sec-array-exotic-objects-defineownproperty-p-desc can invoke it upon a Symbol, for which it would not be defined?

Something like an IsArrayIndex abstract operation would be great.

[anba]

Array index is more formally defined in https://tc39.github.io/ecma262/#sec-object-type.

[annevk]

That still only defines it for Strings as far as I can tell.

[annevk]

For the couple of uses of integer and array index it seems using abstract operations would be much clearer overall.

[rwaldron]

From https://tc39.github.io/ecma262/#sec-object-type:

An array index is an integer index whose numeric value i is in the range +0 ≤ i < 232-1.

Did you want something more explicit, eg. with an esid of its own?

[annevk]

Basically something where https://tc39.github.io/ecma262/#sec-array-exotic-objects-defineownproperty-p-desc step 3 doesn't have a logic error when P is a Symbol and ideally something where you can pass P (even when being a Symbol) to an abstract operation IsArrayIndex() that handles all these details and returns either true or false.

[anba]

I still don't see where 9.4.2.1 contains a logic error. Array indices are a subset of integer indices, and integer indices are a subset of String-valued property keys. So every array index is a String-valued property key.

[annevk]

You're right, I read it the wrong way. Still, I think an abstract operation would be much clearer.